Samples are outside the support for DiscreteUniform distribution

Hello,

I noticed that samples have value outside the support for DiscreteUniform distribution. Here is a simple reproducible example:

import jax.random
import numpyro

import numpyro.distributions as dist

from numpyro.infer import HMC, MCMC, MixedHMC


def model():
    x = numpyro.sample("x", dist.DiscreteUniform(1, 2))


num_samples = 10
kernel = HMC(model, trajectory_length=1.2)
kernel = MixedHMC(kernel, num_discrete_updates=20)
mcmc = MCMC(kernel, num_warmup=1000, num_samples=num_samples, progress_bar=False)
key = jax.random.PRNGKey(0)
mcmc.run(key)
samples = mcmc.get_samples()

print(samples)

Which outputs:

{'x': Array([1, 1, 0, 0, 0, 0, 0, 0, 0, 1], dtype=int32)}

I was expecting values of x to be in [1,2].

Am I using it wrongly or is it a real bug?

Thank you very much for your help.

Start by running the reproducible DiscreteUniform and MixedHMC example from the issue, then inspect the DiscreteUniform and MixedHMC entry points to trace how sampled values are generated. Done means samples remain within the declared [1, 2] support and a regression test covers this case.